OpenVZ Forum


Home » Mailing lists » Users » glusterfs/moosefs/etc as openvz backend?
glusterfs/moosefs/etc as openvz backend? [message #42565] Fri, 29 April 2011 10:46 Go to next message
Aleksandar Ivanisevic is currently offline  Aleksandar Ivanisevic
Messages: 34
Registered: April 2011
Member
From: *parallels.com
Hi,

Is it possible to use a distributed file system like gluster or
moosefs as a backend for openvz? I don't mean mounting it in a VE, but
running the complete VE on it, i.e. mounting /vz/root on a HN on it.

I've googled a bit and all I see is problems, hangs, etc. Has anyone
successfuly done this?
Re: glusterfs/moosefs/etc as openvz backend? [message #42569 is a reply to message #42565] Fri, 29 April 2011 17:36 Go to previous messageGo to next message
Toby Burress is currently offline  Toby Burress
Messages: 2
Registered: April 2011
Junior Member
From: *parallels.com
On Fri, Apr 29, 2011 at 12:46:56PM +0200, Aleksandar Ivanisevic wrote:
>
> Hi,
>
> Is it possible to use a distributed file system like gluster or
> moosefs as a backend for openvz? I don't mean mounting it in a VE, but
> running the complete VE on it, i.e. mounting /vz/root on a HN on it.
>
> I've googled a bit and all I see is problems, hangs, etc. Has anyone
> successfuly done this?

Our disaster recovery scenario, which has been invoked three or four
times, involves running a VE via NFS until a local copy exists, and it
works pretty well. However, they've never run like that for more than
about a day.
Re: glusterfs/moosefs/etc as openvz backend? [message #42570 is a reply to message #42569] Fri, 29 April 2011 17:54 Go to previous messageGo to next message
Angel Bosch Mora is currently offline  Angel Bosch Mora
Messages: 2
Registered: April 2011
Junior Member
From: *127.75.76.dyn.user.ono.com
> Our disaster recovery scenario, which has been invoked three or four
> times, involves running a VE via NFS until a local copy exists, and it
> works pretty well. However, they've never run like that for more than
> about a day.
>

can you please develop a little bit more this setup?
Re: glusterfs/moosefs/etc as openvz backend? [message #42571 is a reply to message #42570] Fri, 29 April 2011 18:11 Go to previous messageGo to next message
Toby Burress is currently offline  Toby Burress
Messages: 2
Registered: April 2011
Junior Member
From: *parallels.com
On Fri, Apr 29, 2011 at 07:54:45PM +0200, Angel Bosch Mora wrote:
> > Our disaster recovery scenario, which has been invoked three or four
> > times, involves running a VE via NFS until a local copy exists, and it
> > works pretty well. However, they've never run like that for more than
> > about a day.
> >
>
> can you please develop a little bit more this setup?

We back up our VEs with rsnapshot. Usually for restores we just mount
the NFS shares for each VE within the VE itself, so that users can grab
their files. However if, as sometimes happens, a RAID controller decides
that it doesn't want to be a RAID controller anymore, and the whole box
goes down, we mount the most recent backup via NFS on another server
and start the VEs there.

You have to disable ACLs, and don't squash root, but other than that it
seems to work fine. It's not a distributed file system, but my guess
is if the file system itself is working properly, a VE will probably
run on top of it with little fuss. There might be issues with POSIX
semantics, though, that these file systems (which I've never used)
may not honor.
Re: glusterfs/moosefs/etc as openvz backend? [message #42577 is a reply to message #42571] Mon, 02 May 2011 07:27 Go to previous message
Aleksandar Ivanisevic is currently offline  Aleksandar Ivanisevic
Messages: 34
Registered: April 2011
Member
From: *parallels.com
Toby Burress <tdb@dimins.com> writes:

> You have to disable ACLs, and don't squash root, but other than that it
> seems to work fine. It's not a distributed file system, but my guess
> is if the file system itself is working properly, a VE will probably
> run on top of it with little fuss. There might be issues with POSIX
> semantics, though, that these file systems (which I've never used)
> may not honor.

Thats what I'm trying to do too, and those were exactly my worries. I
know that it might work with NFS, but I want something more flexible,
so I'm looking into distributed filesystems.

The problem is that I'm not satisfied with daily or hourly rsnapshot
backup, I want to have the replica as close to the original as
possible. We now replicate using DRBD but it introduces high latency
and it becomes really messy once you have more than 2 nodes (ganeti
might help with this, but it was pretty imature on RHEL last time I
looked).

One alternative is also lsyncd but the real deal would be a real
distributed file system where everything happens in the background.
Previous Topic: Switching from VMWare to OpenVZ?
Next Topic: HELP:How can one VM interrupt another running VM ??
Goto Forum:
  


Current Time: Fri Nov 16 06:23:34 GMT 2018